'Big Data' classification using evolutionary computation

Kinny, Daniel (2016) 'Big Data' classification using evolutionary computation. BSc dissertation, University of Portsmouth.

[img] PDF
Restricted to Registered users only

Download (7MB)

    Abstract

    This project aims to assess the suitability of three Evolutionary computational algorithms for the purposes of classifying 'Big Data', from the perspectives of both Algorithmic efficiency in terms of time required, and algorithmic effectiveness in terms of obtainable accuracy. The three algorithms being used to this end are a genetic algorithm, differential evolution, and simulated annealing. The dataset being operated on for these purposes is the 'Connect-4' dataset which is freely obtainable from the UCI Machine Learning Repository. Simulation results have shown Differential Evolution to be a suitable and robust method, capable of very good Convergence and able to attain very high classification accuracy. It has also responded reasonably well to the significant computational demands of working with such a large dataset. The central recommendation of this report is therefore the use or incorporation of differential evolution methods, particularly if future work aims to use complex hybridised evolutionary algorithms for the task of classification, although it is possible this result may only be pertinent to the dataset at hand.

    Item Type: Dissertation
    Departments/Research Groups: Faculty of Technology > School of Computing
    Depositing User: Jane Polwin
    Date Deposited: 05 Aug 2016 12:22
    Last Modified: 05 Aug 2016 12:22
    URI: http://eprints.port.ac.uk/id/eprint/21421

    Actions (login required)

    View Item

    Document Downloads

    More statistics for this item...